Electric shovel positioning method based on density clustering

Abstract

The present invention belongs to the field of open-pit mine production technology, and in particular, relates to a method for positioning an electric shovel based on density clustering. The method uses the scheduling work plan of each shift, GPS dynamic positioning data, and the operation time and trajectory of the shovel, rock crushing tipping point, and ore crushing tipping point of the transport vehicle to calculate the vehicle-shovel correspondence, and uses the DBSCAND algorithm to obtain the preliminary positioning information of the electric shovel. The GPR algorithm is then used to predict the trajectory of the failure position of the electric shovel and the transport vehicle, analyze the data of the vanishing point of the electric shovel signal and the vanishing point of the transport vehicle signal, and perform least squares curve fitting on multiple predicted values. Each intersection of the predicted trajectory is mapped on the line connecting the failure position of the electric shovel and the transport vehicle, and the point with the maximum mapping density is taken as the accurate operating position of the electric shovel. Based on the existing equipment, this method uses a suitable algorithm for calculation, and can obtain the accurate position of the electric shovel without the need for reinvestment in equipment, providing reference data for production scheduling and further realizing lean production in mines.

Description

Technical Field

[0001] The invention belongs to the technical field of open-pit mine production, and in particular relates to an electric shovel positioning method based on density clustering. Background Art

[0002] Open-pit mine is a large-scale production system with mining as the center and transportation as the link. The completion of production indicators and tasks of open-pit mines, as well as the organization and implementation of production processes are all completed through the real-time dispatch of mining and transportation equipment. Therefore, as an important mining and transportation equipment, it is crucial to accurately position the electric shovel.

[0003] In the actual production process, the location of the electric shovel may be physically blocked and electromagnetically interfered. Even if the communication jump technology is used for data transmission, some electric shovels are still in absolute blind spots and the specific location of the electric shovel cannot be obtained. The transport vehicle is a mobile device. It can transmit GPS positioning signals when running near a base station with a good network, but there are also communication blind spots. By using existing resources (i.e. the scheduling work plan for each shift, GPS positioning data, the shovel where the transport vehicle is located, the rock crushing tipping point, and the ore crushing tipping point) and appropriate methods, the accurate positioning of the electric shovel can be obtained, which can further achieve lean production on the basis of reducing equipment investment.

[0004] The Density Based Spatial Clustering of Application with Noise (DBSCAN) algorithm clusters data based on the sparsity of the data. It can find data sets of any shape and size even in noisy data and can also separate outliers. Based on known data, this algorithm can be used for the preliminary positioning of electric shovels.

[0005] Gaussian Process Regression (GPR) is a machine learning method based on Gaussian random functions that has been developed in recent years. It mainly uses sample data to capture the distribution of the entire value function. It has good adaptability to complex problems such as small samples, high dimensions, and nonlinearity, and has strong generalization ability. It can be used for accurate positioning of electric shovels. On the basis of the preliminary positioning of the electric shovel, through the historical data of the electric shovel and the transport vehicle before the GPS failure, GPR is used to predict the running trajectory of the electric shovel and multiple transport vehicles, so as to achieve accurate positioning of the electric shovel.

[0006] The electric shovel can be accurately positioned by using a method combining Gaussian process regression and density clustering (GPR-DBSCAND). Summary of the invention

[0007] The purpose of the present invention is to design an electric shovel positioning method based on density clustering, aiming to use existing data resources, density clustering and Gaussian process regression to achieve accurate positioning of the electric shovel.

[0008] The objective of the present invention is achieved through the following technical solutions:

[0009] The electric shovel positioning method based on density clustering of the present invention is characterized by comprising:

[0010] By utilizing existing resources, i.e. the scheduling work plan for each shift, GPS positioning data, the operating time of the transport vehicle's shovel, rock crushing tipping point, and ore crushing tipping point, the density clustering algorithm is used to cluster the data according to its sparseness, separate the outliers, and determine the initial position of the shovel;

[0011] Based on the preliminary positioning of the electric shovel, the historical data of the electric shovel and the transport vehicle before GPS failure are used to predict the running trajectories of the electric shovel and multiple transport vehicles in the blind area using Gaussian process regression to achieve accurate positioning of the electric shovel.

[0012] The specific steps of the electric shovel positioning method based on density clustering include:

[0013] Step 1: Dynamically collect GPS data. The system calculates the data to prevent drift and filters the GPS drift points through the characteristics of the time series. According to the correction algorithm proposed by the time series theory, the original sequence has irregular points of GPS positioning drift through white noise test, and the sample points of the local test sequence are corrected based on the search of the autocorrelation coefficient.

[0014] Step 2: Reading of vehicle and shovel planning data. The dispatch plan is issued, including the shovel number, operating status, operation type, operation direction, unloading target and specific transport vehicle number, which provides the prerequisite for the calculation model;

[0015] Step 3: Automatically analyze the route of the transport vehicle and the location of the shovel. According to the scheduling plan, first summarize the location of the shovel according to the running track of each transport vehicle, and preliminarily determine the starting point of the transport vehicle loading and the destination of unloading according to the route; after 1-2 cycles, plan the route of the transport vehicle and the location range of the shovel;

[0016] Step 4: Prediction of the failure trajectory of transport vehicles. Determine the last disappearance location of multiple transport vehicles’ GPS, combine the approximate location of the shovel determined by density clustering, and use Gaussian process regression to predict the GPS failure trajectory;

[0017] Step 5: Prediction of the failure trajectory of the electric shovel: predicting the running trajectory of the electric shovel based on the running trajectory before the electric shovel signal disappears;

[0018] Step 6: Accurately locate the shovel operation position in the absolute blind spot, perform data analysis on the shovel signal vanishing point and the transport vehicle signal vanishing point, perform least squares curve fitting on multiple predicted values, map each intersection point on the line connecting the failure positions of the shovel and the transport vehicle, and take the point with the maximum mapping density as the final solution point, that is, determine the shovel operation position.

[0019] Advantages of the present invention:

[0020] The electric shovel positioning method based on density clustering of the present invention uses the scheduling work plan of each shift, GPS dynamic positioning data, and the operation time and trajectory of the shovel, rock crushing tipping point, and ore crushing tipping point of the transport vehicle to calculate the vehicle-shovel correspondence, and uses the DBSCAND algorithm to obtain the preliminary positioning information of the electric shovel. The GPR algorithm is then used to predict the trajectory of the failure position of the electric shovel and the transport vehicle to obtain the accurate position of the electric shovel. Based on the existing equipment, this method uses a suitable algorithm for calculation, and can obtain the accurate position of the electric shovel without the need for reinvestment in equipment, providing reference data for production scheduling and further realizing lean production in mines. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 The figure is a flow chart of the electric shovel positioning method based on GPR-DBSCAND of the present invention.

[0022] Figure 2 This is a diagram of the GPR-based accurate positioning method of the present invention.

[0023] Figure 3 This is the actual positioning diagram of the electric shovel based on GPR-DBSCAND of the present invention. DETAILED DESCRIPTION

[0024] The specific implementation of the present invention is further described below in conjunction with the accompanying drawings.

[0025] like Figure 1-3 As shown, the electric shovel positioning method based on density clustering of the present invention is characterized by comprising:

[0026] Using the plan of two shifts in a working day (night shift: 18:30-07:30, day shift: 07:30-18:30), GPS is used to locate the shovel, and the workload of the vehicle and the type of ore it transports are calculated based on the time the vehicle is at the shovel, rock breaking point, and ore breaking point. The variables used in the model are defined as follows:

[0027] Transport vehicle: k i ∈K=(k0,k1,k2…k n );

[0028] There are multiple vehicles working every day: k i =(Id k , loadi );

[0029] Transport vehicle GPS information: g ki ∈G=([k0,t0,B0,L0],…[k n , t i , B n , L n ]);

[0030] Ore type: t i ∈T=(rock,ore0,ore1...,ore l );

[0031] GPS information of rock breaking point: R gps ;

[0032] GPS information of mining point: O gps ;

[0033] Electric shovel: e i ∈E=(e0,e1…e m );

[0034] There are many electric shovels working every day: i =(Id e , B i , L i , t i );

[0035] Production plan: p i ∈P=(p0,p1,…p j );

[0036] Multiple plans can be issued every day, each plan will include the time of the plan issuance, transportation to a specific shovel work information: p = (Id p ,Time start ,Time end ,(e i :(k1,k2,…k n )));

[0037] The goal of this model is to calculate the positions of all blind-spot shovels and the routes of transport vehicles in both shifts on that day:

[0038] L(G,P)

[0039] F tstart，tend (g ki , R gps O gps , p k , L(G,P))

[0040] The L(G, P) function analyzes the GPS data and plans of all vehicles in the class and calculates the position of all shovels working on the day (B i , L i ) to make a preliminary judgment and use the density clustering algorithm to mark the preliminary position of the shovel.

[0041] The algorithm is as follows:

[0042]

[0043]

[0044] After obtaining the preliminary information of the positions of all the working shovels on that day through L(G, P), tstart，tend (g ki , R gps O gps , p k , L(G, P)) can obtain the entire transportation path of each operating vehicle on the day. The algorithm is as follows:

[0045]

[0046] After obtaining the preliminary position of the shovel and the driving route of the transport vehicle, further accurate positioning is carried out.

[0047] like Figure 2 As shown in the figure, based on historical data, point A is determined as the location where the electric shovel signal fails, which is expressed as:

[0048]

[0049] in, is the latitude, is the longitude, is the height, For speed.

[0050] According to the failure positions of different transport vehicles, the failure position B of the transport vehicle is calculated. First, the failure position matrix of the transport vehicle is established, and the formula is

[0051]

[0052] Where B, L, H, and V are latitude, longitude, altitude, and speed, respectively. Each line is Ψ i , n is the number of transport vehicles.

[0053] Point G is the precise working position of the shovel obtained by solving the known data. The formula is:

[0054]

[0055] Solve the average latitude position of n sampling points, the formula is:

[0056]

[0057] In the formula, represents the average value of latitude, B i is the latitude value of each transport vehicle.

[0058] Solve the longitude position average of n sampling points, the formula is:

[0059]

[0060] In the formula, Indicates the average value of longitude, L i is the latitude value of each transport vehicle.

[0061] To solve the average height position of n sampling points, the formula is:

[0062]

[0063] In the formula, Indicates the average height, H i is the height of each transport vehicle.

[0064] Similarly,

[0065]

[0066] Get the mean point of the transport vehicle failure position,

[0067]

[0068] Solve the deviation of each transport vehicle’s position from the mean point:

[0069]

[0070] In the formula, k is the amplification factor, which improves the positioning accuracy.

[0071] Solve for the mean square error for each transport vehicle:

[0072]

[0073] The position of the transport vehicle is processed according to the deviation and mean square error values, and serious drift points and false value points are eliminated. In order to introduce the statistical law, the mean square error adjustment factor p∈(0,1) is used. If the formula is satisfied

[0074] ε i ≤p*δ

[0075] Then the point is retained, otherwise it is removed.

[0076] Finally, the dynamic and precise positioning of the transport vehicle's GPS signal failure position is obtained as follows:

[0077] Ψ v =[B v L v H v V v ]

[0078] Figure 2 Medium T e is the prediction of the failure position trajectory of the electric shovel based on Gaussian process regression, T v1 、T v2 …T vn are the failure position trajectory predictions of n transport vehicles, T vo Predict the precise failure location trajectory of the transport vehicle.

[0079] Using the Gaussian process model to solve the prediction problem, given a training set C, where (x i ,y i )∈P γ , are γ-dimensional GPS input and output data respectively. N is the training sample size. Define the following model:

[0080] y i =f(x i )+ε

[0081] Where f is the function value, x i is the input vector, ε is the mean value of 0, and the variance is Gaussian noise.

[0082] The noise error is expressed in the covariance function as follows:

[0083]

[0084] Wherein, δ(x, x) is the Krone function, and its value is 1 if and only if x=x, and 0 otherwise.

[0085] Key assumptions based on the Gaussian process model: observed values ​​y and predicted values ​​y * It can be regarded as a point sampled from a joint Gaussian distribution, so the prior distribution of the observed value y can be obtained.

[0086]

[0087] The observed and predicted values ​​are distributed as:

[0088]

[0089] Among them, K(X, X), K(X, x * ), K(x * , X), k(x * , x *) are covariance matrices, K(X, X) = K N =(k ij ) is a covariance matrix of N*N order symmetric positive definite, and the matrix element k ij = k(x i , x j ) represents x i With x j The correlation between them.

[0090] K(X,x * )=K(x * , X) T For the test matrix point x * The covariance matrix between the training set X, k(x * , x * ) is the test point x * Its own covariance; I n is the N*N unit matrix. The specific form is as follows:

[0091]

[0092] K(x * , X) = [k(x * , x1)k(x * , x2)…k(x * , x n )]

[0093] Therefore, the predicted value y can be obtained * The posterior distribution is:

[0094] y * |X,y,x * ~N(μ(y * ), cov(y * ))

[0095] Where μ(y * ) is the test point x * The corresponding mean, cov(y * ) is the corresponding variance:

[0096]

[0097] Gaussian distribution is also a normal distribution. Its mean has the highest probability. The farther it deviates, the smaller the probability of the value taken, that is, the greater the error. Therefore, the mean μ(y * ) is the predicted value.

[0098] The least squares curve fitting of multiple predicted values ​​is used to obtain T e and T vo , T v1、T v2 …T vn .

[0099] T e With T vo 、T v1 、T v2 …T vn Different intersection points will appear. Map each intersection point on the AB line and take the point with the largest mapping density as the final solution point, that is, determine the shovel operating position G.

Claims

1. The electric shovel positioning method based on density clustering is characterized by include: By utilizing existing resources, including the scheduling work plan for each shift, GPS positioning data, the operating time of the transport vehicle's shovel, rock crushing tipping point, and ore crushing tipping point, the density clustering algorithm is used to cluster the data according to its sparsity, separate the outliers, and determine the initial position of the shovel; Based on the preliminary positioning of the electric shovel, the historical data of the electric shovel and the transport vehicle before GPS failure are used to predict the running trajectories of the electric shovel and multiple transport vehicles in the blind area using Gaussian process regression to achieve accurate positioning of the electric shovel. The specific steps of the electric shovel positioning method based on density clustering include: Step 1: Dynamically collect GPS data; the system calculates the data to prevent drift and filters the GPS drift points through the characteristics of the time series; according to the correction algorithm proposed by the time series theory, the original sequence has irregular points of GPS positioning drift through white noise test, and the sample points of the local test sequence are corrected based on the search of the autocorrelation coefficient; Step 2: Read the vehicle and shovel planning data; issue the scheduling plan, including the shovel number, operating status, operation type, operation direction, unloading target and specific transport vehicle number, which provides the prerequisite for the calculation model; Step 3: Automatically analyze the route of the transport vehicle and the location of the shovel for preliminary calculation; according to the scheduling plan, first summarize the location of the shovel according to the running track of each transport vehicle, and preliminarily determine the starting point of the transport vehicle loading, the shovel and the destination of unloading according to the route; after 1-2 cycles, plan the route of the transport vehicle and the location range of the shovel; Step 4: Prediction of the failure trajectory of transport vehicles: Determine the last disappearance location of the GPS of multiple transport vehicles, combine the approximate location of the electric shovel determined by density clustering, and use Gaussian process regression to predict the GPS failure trajectory; Step 5: Prediction of the failure trajectory of the electric shovel: predicting the running trajectory of the electric shovel by the running trajectory before the electric shovel signal disappears; Step 6: Accurately locate the shovel operation position in the absolute blind spot, perform data analysis on the shovel signal vanishing point and the transport vehicle signal vanishing point, perform least squares curve fitting on multiple predicted values, and the shovel and transport vehicle trajectory predictions will have different intersection points. Map each intersection point on the line connecting the failure positions of the shovel and the transport vehicle, take the point with the largest mapping density as the final solution point, and determine the shovel operation position.

Images